[peiji1984]

Hello everyone!

I am doing a FSI simulation for a rotating impeller with MFX bench. The deformation of the impeller is kind of small(0.2-0.5mm) according to experiment. For the structure side, I set large deformation option to off in workbench, therefore, no nonlinear solution will be obtained. For one way coupling, the result seems similar to experiment results. But when two way coupling is used, the amplitude of deformation is much smaller like 0.002mm. Also, the interface load convergence will take very long time for a time step(more than a hunderd coupling/stagger iterations).
I also try to open the large deformation option(nonlinear solution) to solve. Although the average amplitude seems right, amplitudes for different time steps of a cycle is not equal.(it should be basically equal according to the experiment) .

I know most FSI problems are nonlinear problems. Many linear questions use one way coupling are sufficient. But anyway can I use two way coupling to analyze it? Do you have any experiences about this?

Thanks in advance!

Craig

[singer1812]

Keep large deformation on. If you keep the large deformation off, the load from the CFX side is applied to the undeformed structure shape, which can lead to errors.

Regarding your amplitudes:

Are you sure you have run out the analysis far enough to reach a quasi-steady solution? You might still be seeing startup effects.

Also

This type of analysis would be very sensitive to modeling inconsistancies. Are you sure your time step is sufficient? Are you sure your mesh is refined enough? Are you sure your BCs are sufficiently far away to not effect solution?

[peiji1984]

Singer1812, thank you for your reply and some suggestions!
1. About keeping deformation to off, what is "CFX load is applied to the undeformed structure shape" mean? Could you give a little more details? I mean, I didn't catch error in my case except unbelievable results (too small deformation).
2. I have run 6 revolutions before fsi calculation, I think it should be steady, but I will try more.
3. There are 120 timesteps in one revolution, I am not sure it is sufficient. I will try.
4. Fluid mesh, I think, is good, but I don't know for FEM mesh which I generated directly in workbench. Any suggestion?
5.When I set large deformation on, the result is probably like B in the figure, but experient and one way coupling results are like A.
1.png

[singer1812]

What I mean by load applied to undeformed shape means exactly that. As the structural shape deformes and forces are applied to the structure, for large deformations turned off, those forces are applied to the structure in the original shape.

For an extreme example, if you have a sloped surface that is originally 45deg off the x plane, the forces on that surface will have a component of 0.707F in the x and y direction.

Now lets say during solution the surface deforms so that it lies entirely inplane with the x axis. With large deformations set to off, the forces on that surface will still have the same components listed above, because it is still applied on the undeformed shape. With large deformations on, it will correctly apply the force on the deformed shape.

So, it is up to you to determine whether this approximation will hold up for your solution.

[peiji1984]

Thanks a lot! singer1812. I was confused by this.
The reason I want to keep large deflection off is that the deflection rotor in my case is small, I think it is linear solution. I didn't know what it means exactly at that moment.
So is it right to use this to solve linear case?